Research/Featured

Verify research you did not do yourself

Maps a memo from its question to its conclusion, shows which step each authority actually supports, names the one proposition that would take the whole thing down, and lists what has to be pulled before anyone relies on it.

About 30 minadvancedLitigation, In-house

Your prompt6,332 characters

Still to fill in: The memo or answer to verify, The question that was put, Where the answer has to work

RoleYou are the lawyer who signs the advice, which means you are the one who finds out whether the research under it holds. You do not read a memo for whether it sounds right. You break it into the propositions it depends on, ask what carries each one, and treat fluency as a warning sign rather than a credential. You know you cannot open a database from here, so you never pretend to have checked whether a case exists.What I needAudit the analysis below, which came from another lawyer in my firm in answer to The question that was put, and tell me whether I can rely on it in Where the answer has to work.InputsThe memo or answer: The memo or answer to verify The question that was put: The question that was put Where the answer has to work: Where the answer has to work Where it came from: another lawyer in my firm What I already hold that cuts against it: What you already know that cuts against itHow to work this1. Restate in one sentence the conclusion the memo actually reaches, in your own words, and say whether it answers The question that was put or a neighbouring question. If it reaches no conclusion, say so and stop the audit there: a survey can be summarized but not verified. 2. Map the dependency chain. Number every step the memo takes from The question that was put to its conclusion, state the proposition each step asserts, and name the authority the memo offers for it. A numbered step with nothing under it is the most important thing in your output. 3. Classify the support for each numbered step into exactly one of five: holds it (the cited authority plainly supports the proposition as stated), holds it with an inference (name the inferential step the memo skipped), does not reach it (the authority is about something adjacent), cannot carry it (a treatise cited for a holding, a trial court cited for a binding rule, dicta cited as a holding, an out-of-forum case cited as controlling), or nothing. 4. Identify the load-bearing proposition: the single step that, if it fails, takes the conclusion with it. Say in one sentence what would have to be true for it to fail, and whether What you already know that cuts against it is that thing. 5. Name what the analysis never addresses. Work through this list explicitly and say "addressed" or "absent" for each: the strongest counterargument, a contrary line of authority, choice of law, preemption, limitations, exhaustion, standing, the standard of review, the procedural posture, and whether the governing text has been amended since the authorities were decided. Absence is the defect that fluent writing hides best. 6. Mark every case, statute, rule, and quotation in the memo [PULL IT] with the reason it matters, and state plainly that you cannot confirm from here that any of them exist, say what they are quoted as saying, or remain good law. Do not attempt to verify any of it from memory, and do not supply a citation the memo does not contain. This is not a citation-form audit; it is an audit of whether the reasoning stands. 7. Close with a reliance verdict in one of four forms: rely on it; rely on it once these specific sources are pulled; do not rely on it; or it answered a different question from the one I asked.Ask me firstYou have the memo, the question, the forum, and the contrary authority I hold. Ask me these four, which the paper cannot tell you, then stop and wait: 1. What turns on this, and can it be undone if the memo is wrong? A term-sheet position and a letter to a regulator deserve different amounts of my suspicion. 2. Does the conclusion match what you already believed before you read it? I will look hardest where your instinct and the memo agree, because agreement is where nobody checks. 3. Has any of this already gone out, and to whom: a client, a court, a counterparty, a board? 4. Can you get to a real database before you have to act, or is the deadline going to force you to rely on this unverified? Do not begin the audit until I answer. If I tell you to proceed anyway, state each assumption at the top of your output and mark it [ASSUMPTION - verify].Output formatOpen with the reliance verdict in one line and the load-bearing proposition in one line. Then the dependency map as a numbered table: Step | Proposition | Authority the memo offers | Classification | Note. Then "What is absent," running the full list from step 5 with addressed or absent against each. Then the pull list: every authority, the proposition it has to support, and what a negative result would do to the conclusion. Then the assumptions the audit itself rests on. Close with one line naming the two of my answers that most changed this audit, and what the verdict would have been without them. If an answer changed nothing, say so, because it means the question was not worth asking.Never do this- If the audit would read the same for any memo on any topic, it is too generic. Every classification quotes the memo's own sentence and its own citation. - No hedging filler. "The analysis is arguably incomplete" is not a finding. Name the step, the proposition, and what is missing under it. Do not tell me to consult an attorney; I am the attorney who has to sign this. - Never confirm that a cited case exists, is quoted accurately, or remains good law, and never supply a citation of your own. Everything the memo cites is [PULL IT], and everything you are tempted to add is [UNVERIFIED - do not use]. - Where you cannot tell whether an authority supports a step, say you do not know and say what would settle it. Do not smooth over the gap with fluent prose, which is the exact failure you are auditing. - Do not pad. A three-step memo gets a three-step map. Length is not value.Before you answer- Did I classify every numbered step, including the ones with nothing under them? - Did I run the full absence list in step 5 and mark each one, rather than listing only what I noticed? - Did I assert that any authority exists, is accurate, or is current? - Is the load-bearing proposition genuinely load-bearing, or just the one I found most interesting? - Would this audit be useless against a different memo? It should be.

The run walks turn one, the pressure test, the follow-up, and a check on what came back. The Cockpit adds driver's-seat tunes. Your values come with either one.

Pressure-test it

Makes the AI switch hats and attack its own answer.

Opposing counsel's brief arrives citing the two authorities this memo never mentions. Read your own audit as the associate who now has to explain to the client why nobody found them: which step of the dependency map should have surfaced them, whether the absence list caught the gap or named a category without looking in it, and what in the audit gave false comfort. Then tell me the one check that would have caught it.

Go deeper

Pushes the work further once the basics are right.

Verification produces a pull list and nothing else until somebody pulls it. Turn the [PULL IT] items into the assignment: each source paired with the proposition it has to support, the exact question to answer once it is on screen, what a negative answer does to the conclusion, and who owns it by when. Order it so the load-bearing proposition gets checked first and the rest can be abandoned if it fails.

4Check what came backPaste the answer here and work a checklist against this prompt's own rules.

Before you run it

What to gather first

The memo exactly as you received it, including its footnotes
The question that was actually asked, in the form it was asked
The jurisdiction and forum the answer has to work in
Where the memo came from, since a model and a senior associate fail differently
Any authority you already know that cuts the other way

Watch for

This audit cannot tell you whether a cited case exists. A memo can pass every structural check here and rest entirely on authority that was invented. Pull the sources.
The model auditing the memo can hallucinate as readily as the model that wrote it. Treat any authority it names, including one it offers as contrary, as unverified.
A memo that reaches the conclusion you wanted is the one you will check least. That is the reason this prompt asks what you already believed.
Courts have sanctioned lawyers for filings built on fabricated authority. A [PULL IT] flag is an assignment to a human, not a disclaimer that moves the risk.
Do not paste privileged memoranda or client-confidential analysis unless your firm's AI policy and the engagement terms permit it.

What comes back

Opens with the reliance verdict and the load-bearing proposition, each in one line. Then a numbered dependency table (step, proposition, authority offered, classification, note), an explicit "What is absent" run through choice of law, preemption, limitations, exhaustion, standing, standard of review, posture, amendment, and the counterargument, then a pull list keyed to propositions, the audit's own assumptions, and the answer-consumption line.

See an example of what you’ll get

*(After I answered: this drives whether we send a cease-and-desist this week, it is reversible but embarrassing; the conclusion is what I expected; nothing has gone out yet; and I can get to Westlaw tomorrow.)* Verdict: do not rely on it yet. The conclusion may well be right, but it rests on one proposition the memo never supports and never flags. Load-bearing proposition: that Colorado's compensation threshold for enforceable non-solicits is measured at the time of signing rather than at the time of enforcement. If it is measured at enforcement, Delgado is outside it and the whole memo inverts. Dependency map | # | Proposition the step asserts | Authority the memo offers | Classification | Note | |---|---|---|---|---| | 1 | Colorado restricts non-solicitation covenants by statute | C.R.S. § 8-2-113 | Holds it | The memo's only citation to operative text. [PULL IT] | | 2 | The statute exempts covenants for workers above a compensation threshold | Same section | Holds it with an inference | The memo quotes the exemption but never quotes the threshold provision itself or states the figure | | 3 | Delgado is above the threshold | Her 2024 offer letter | Cannot carry it | An offer letter establishes what she was offered, not what she earned in the measuring year. Nothing in the memo is a payroll record | | 4 | The threshold is measured at signing | Nothing | Nothing | This is the step the conclusion turns on and the memo does not argue it, cite it, or notice it | | 5 | A Colorado court would apply Colorado law despite her move to Denver from out of state | "Colorado has the most significant relationship" | Does not reach it | The memo cites a choice-of-law case about tort claims. She moved within Colorado in any event, which the memo never says, so the question may not arise | | 6 | The covenant's geographic scope is reasonable | Two federal district decisions | Cannot carry it | Both are federal trial courts predicting state law. Persuasive at best in state court, and the memo presents them as settled | What is absent | Issue | Status | |---|---| | The strongest counterargument | Absent. The memo never states Delgado's best position | | A contrary line of authority | Absent. Every citation runs one way, which for a contested statute is itself a signal | | Choice of law | Addressed, but at step 5, badly | | Preemption | Absent, and probably not in play | | Limitations | Absent. When the covenant was breached is never discussed | | Exhaustion | Not applicable | | Standing | Not applicable | | Standard of review | Absent. A preliminary injunction standard governs what you can actually get, and the memo answers enforceability in the abstract | | Procedural posture | Absent. The memo answers "is it enforceable," and you asked what you can do this week | | Amendment since the authorities | Absent and serious. You told me a 2023 amendment raised the threshold. The memo does not mention an amendment, and its two federal cases predate it if they are the vintage they appear to be | Pull list 1. C.R.S. § 8-2-113, current text with effective dates. Has to support steps 1, 2, and 4. If the threshold provision measures compensation at enforcement rather than at signing, the conclusion reverses. Pull this first; everything else can wait on it. 2. The 2023 amendment you mentioned. Has to establish which version applies to a covenant signed in 2024 and enforced now. 3. The two federal decisions at step 6. Have to support geographic reasonableness. Confirm they exist, that they apply Colorado law, and that they postdate the amendment. 4. The choice-of-law case at step 5. Confirm what claim type it involved. 5. Delgado's payroll records for the measuring year. Not a legal source, and it is the only thing that establishes step 3. Assumptions this audit rests on. That the memo's quotation of the exemption language is accurate [verify: I cannot confirm any quotation]. That "the 2023 amendment" you mentioned is to the same section [verify]. That the two federal decisions are district court decisions, inferred from their citation format as the memo presents them [verify]. What your answers changed. That you already believed the conclusion is why step 4 is the headline rather than a footnote. An analysis that confirms a prior belief gets read for whether it sounds right, and step 4 sounds entirely right; it is simply not there. Without that answer I would have led with step 6, which is the more visible defect and the less dangerous one. Your answer that you can reach a database tomorrow is why this output is a ranked pull list with a first item rather than a recommendation to send nothing. Your answer that nothing has gone out yet changed nothing in the audit itself, though it is the reason there is no remediation section, so I should not have asked it.

Why this prompt is built the way it is

## Framework 1. **State the conclusion the memo actually reaches,** in one sentence. A memo that surveys without concluding cannot be verified, only summarized. 2. **Map the dependency chain** from question to conclusion. Number each step and name the authority cited for it. A step with no authority is the finding. 3. **Classify the support for each step:** the authority plainly holds it, it holds it after an inferential step you must name, it does not reach it, it cannot carry the proposition by its nature, or there is nothing. 4. **Find the load-bearing proposition,** the one step that takes the conclusion with it, and say what would have to be true for it to fail. 5. **Name the absences:** the counterargument, the contrary line, the threshold issue, the amendment, the posture that changes the standard. 6. **Mark every citation [PULL IT].** You cannot verify that a case exists, says what it is quoted as saying, or is still good law. Say so plainly. 7. **Give a reliance verdict,** not a summary.